tf_executor.fetch %0#2 : !tf_executor.control
  }
  func.return
}

// CHECK-LABEL: func @testNextIterationMissingSourceDevice
func.func @testNextIterationMissingSourceDevice() {
  tf_executor.graph {
    // CHECK:      tf_executor.NextIteration.Source
    %0:3 = tf_executor.NextIteration.Source : tensor<i64> {T = "tfdtype$DT_INT64"}

    %outputs_1, %control_1 = tf_executor.island wraps "tf.TPUReplicatedInput"(%outputs_0) {N = 1 : i64, T = "tfdtype$DT_FLOAT", device = "", name = "input1"} : (tensor<0xf32>) -> tensor<0xf32>

//   %cst_2 = "tf.Const"() {value = dense<[0.3, 0.4, 0.3, 0.4, 0.3, 0.4, 0.3, 0.4]> : tensor<8xf32>} : () -> tensor<8xf32>

  %0 = arith.constant dense<2.000000e+00> : tensor<f32>
  %1 = arith.constant dense<1.000000e+00> : tensor<f32>
  %2 = "tf.AddV2"(%arg0, %1) {T = "tfdtype$DT_FLOAT", device = "", name = "StatefulPartitionedCall/add"} : (tensor<1xf32>, tensor<f32>) -> tensor<1xf32>
  %3 = "tf.Mul"(%2, %0) {T = "tfdtype$DT_FLOAT", device = "", name = "output_node"} : (tensor<1xf32>, tensor<f32>) -> tensor<1xf32>
  func.return %3 : tensor<1xf32>

// CHECK-NEXT: %[[Y:[a-z0-9]*]], {{.*}}_FusedBatchNormEx
// CHECK-NEXT: return %[[Y]]

%outputs_6, %outputs, %outputs_0) {Tdense = ["tfdtype$DT_FLOAT", "tfdtype$DT_FLOAT"], dense_shapes = [#tf_type.shape<>, #tf_type.shape<>], device = "", num_sparse = 2 : i64, ragged_split_types = [], ragged_value_types = [], resultSegmentSizes = array<i32: 2, 2, 2, 2, 0, 0>, sparse_types = ["tfdtype$DT_STRING", "tfdtype$DT_INT64"]} : (tensor<32x!tf_type.string>, tensor<0x!tf_type.string>, tensor<2x!tf_type.string>, tensor<2x!tf_type.string>, tensor<0x!tf_type.string>, tensor<0xf32>, tensor<0xf32>) ->...

  attributes  {tf.entry_function = {inputs = "args_0,args_1", outputs = "rets_0,rets_1"}} {
    %0:2 = tf_executor.graph {
      %1:3 = tf_executor.island wraps "tf.IdentityN"(%arg0, %arg1) {T = ["tfdtype$DT_FLOAT", "tfdtype$DT_INT32"], device = "", name = "identity"} : (tensor<*xf32>, tensor<2x4x6x8xi32>) -> (tensor<*xf32>, tensor<2x4x6x8xi32>)
      tf_executor.fetch %1#0, %1#1 : tensor<*xf32>, tensor<2x4x6x8xi32>
    }

  %0 = "tf.Conv2D"(%arg, %filter) {T = "tfdtype$DT_FLOAT", data_format = "NHWC", dilations = [1, 2, 3, 1], padding = "SAME", strides = [1, 4, 5, 1]} : (tensor<256x32x32x3xf32>, tensor<3x3x3x16xf32>) -> tensor<256x8x7x16xf32>
  %1 = "tf.BiasAdd"(%0, %bias) {T = "tfdtype$DT_FLOAT", data_format = "NHWC"}: (tensor<256x8x7x16xf32>, tensor<16xf32>) -> tensor<256x8x7x16xf32>

    // CHECK-NEXT:      type: DT_FLOAT
    // CHECK-NEXT:    }
    // CHECK-NEXT:  }
    %0:2 = tf_executor.island wraps "tf.Placeholder"() {name = "dummy", dtype = "tfdtype$DT_FLOAT", emptylist = [], typelist = ["tfdtype$DT_INT32", "tfdtype$DT_FLOAT"]} : () -> tensor<*xi32>
    tf_executor.fetch
  }
  func.return

    tf_executor.fetch
  }
  func.return
}

// CHECK-DAG: name: "foo"
func.func @foo() -> tensor<10xf32> {
  %0 = tf_executor.graph {